The advertising industry tends to move in cycles. A new technology emerges, the market experiments, and standards bodies eventually step in to codify how it should work.
That pattern made sense when innovation moved at the speed of APIs, but it may not make sense in the age of AI.
Over the past year, we’ve seen a wave of efforts to standardize agent workflows, from Model Context Protocol (MCP) to emerging agent-to-platform compliance frameworks designed to formalize how AI systems interact with software.
Organizations like the IAB Tech Lab are already moving in this direction. Their proposed Agent Registry is an early attempt to introduce accountability and transparency into a world where software agents may increasingly act on behalf of advertisers.
Efforts like this are valuable. They acknowledge that agents will become real economic participants in the advertising ecosystem.
But they also raise a deeper question: should the industry focus on standardizing agent infrastructure, or on empowering agents to navigate the infrastructure that already exists?
At this early juncture, maintaining flexibility may be the more important priority.
AI Is Moving Too Fast for Protocols
Most emerging agent protocols assume a structured world: predefined objects, mapped schemas, and deterministic workflow steps.
That’s how APIs behave. But frontier AI models are adaptive generalists.
In the past six months alone, agents have evolved from executing defined tasks to interpreting intent, adjusting logic dynamically, and navigating platform constraints in real time.
If that trajectory continues, rigid interoperability frameworks may matter far less than we expect.
The real question is whether advertisers are actually asking for protocol-level alignment, or whether they’re asking for outcomes, control, and simplicity.
In a world where intelligence is compounding every quarter, locking in process standards too early risks ossifying workflows before we understand what agents will ultimately become.
Agents That Learn Systems
Instead of waiting for universal protocol alignment, sophisticated teams are training agents with platform-specific skills.
Rather than forcing every platform to expose identical schemas, these agents learn how existing systems already work — how to navigate The Trade Desk, Google Ads, Amazon, retail media APIs, and proprietary pipes the same way a human operator would.
In this model, an agent doesn’t rely on a rigid protocol layer.
It carries a library of operational skills — modular capabilities that allow it to interact with different systems, interpret their interfaces, and execute tasks across them.
In other words, the focus moves from standardizing the plumbing to training the intelligence.
The future of integration may not be standardized software interfaces — it may be intelligent systems that simply learn how the ecosystem works.
What advertisers actually want is simple:
- Set a budget at a normalized layer.
- Define objectives once.
- Deploy across platforms.
- Update rules automatically.
- Measure outcomes consistently.
No CMO has ever asked for a standardized agentic compliance protocol.
They’ve asked for abstraction, orchestration, and less operational burden.
If agents can interpret campaign intent and dynamically translate it into platform-specific execution, then the true unlock is an intent layer that adapts to wherever media lives.
That’s a very different future than the one most standards conversations assume.
The System Doesn’t Need to Change Yet
The current moment can feel chaotic. New agent frameworks are emerging, competing specifications are being proposed, and model capabilities are evolving faster than most documentation can keep up.
For an industry accustomed to carefully governed standards, that uncertainty can be uncomfortable. But early markets rarely begin with perfect coordination.
The web itself is proof of this. Systems were built independently, interfaces evolved organically, and interoperability emerged over time as the best approaches gained adoption.
AI agents may follow a similar path.
Instead of requiring every platform to adopt a new protocol layer immediately, intelligent systems can operate natively across the web as it already exists — navigating interfaces, interacting with APIs, and orchestrating workflows across disparate platforms without requiring fundamental changes to underlying infrastructure.
In this model, intelligence adapts to systems rather than forcing systems to adapt to intelligence.
That doesn’t mean standards are unnecessary. Over time, organizations like the IAB Tech Lab will play an important role in formalizing best practices — creating registries, governance models, and interoperability frameworks once the ecosystem has matured and real operational patterns have emerged.
But those standards should be informed by how agents actually operate in production environments, not imposed before those patterns are understood.
What will ultimately matter isn’t which protocol wins. It’s which systems deliver measurable outcomes and maintain clean feedback loops.
In an agent-driven world, advantage will accrue to whoever controls the signal — the normalized data, real-world performance inputs, and outcome measurement that allow intelligent systems to continuously improve.
Premature alignment risks freezing the industry around assumptions that may not hold six months from now.
